Performance of gene-expression profiling test score variability to predict future clinical events in heart transplant recipients

Abstract

Background: A single non-invasive gene expression profiling (GEP) test (AlloMap®) is often used to discriminate if a heart transplant recipient is at a low risk of acute cellular rejection at time of testing. In a randomized trial, use of the test (a GEP score from 0-40) has been shown to be non-inferior to a routine endomyocardial biopsy for surveillance after heart transplantation in selected low-risk patients with respect to clinical outcomes. Recently, it was suggested that the within-patient variability of consecutive GEP scores may be used to independently predict future clinical events; however, future studies were recommended. Here we performed an analysis of an independent patient population to determine the prognostic utility of within-patient variability of GEP scores in predicting future clinical events.

Methods: We defined the GEP score variability as the standard deviation of four GEP scores collected ≥315 days post-transplantation. Of the 737 patients from the Cardiac Allograft Rejection Gene Expression Observational (CARGO) II trial, 36 were assigned to the composite event group (death, re-transplantation or graft failure ≥315 days post-transplantation and within 3 years of the final GEP test) and 55 were assigned to the control group (non-event patients). In this case-controlled study, the performance of GEP score variability to predict future events was evaluated by the area under the receiver operator characteristics curve (AUC ROC). The negative predictive values (NPV) and positive predictive values (PPV) including 95 % confidence intervals (CI) of GEP score variability were calculated.

Results: The estimated prevalence of events was 17 %. Events occurred at a median of 391 (inter-quartile range 376) days after the final GEP test. The GEP variability AUC ROC for the prediction of a composite event was 0.72 (95 % CI 0.6-0.8). The NPV for GEP score variability of 0.6 was 97 % (95 % CI 91.4-100.0); the PPV for GEP score variability of 1.5 was 35.4 % (95 % CI 13.5-75.8).

Conclusion: In heart transplant recipients, a GEP score variability may be used to predict the probability that a composite event will occur within 3 years after the last GEP score.

Trial registration: Clinicaltrials.gov identifier NCT00761787.

Figures

Fig. 1

Timing of four serial gene expression profiling scores to predict future clinical events. Four gene expression profiling (GEP) scores were collected beginning day 315 post-transplant. The first and the fourth GEP score were at least 85 days apart and at most 780 days apart. Clinical events were observed ≥ 3 years after the last available score (GEP 4). Therefore, the overall follow-up was up to 6 years (if the patient enrolled in the beginning of 2006 and had a clinical follow-up in late 2011). DOT: date of transplant

Fig. 2

Creation of study cohort using the Cardiac Allograft Rejection Gene Expression Observational (CARGO) II study database. Of 737 patients enrolled in CARGO II study, 171 had at least two gene-expression profiling (GEP) scores and three year follow-up data available. For our cohort, we finally selected 91 patients with at least four serial gene expression profiling scores. These patients were assigned to either an event group or a control group

Fig. 3

Distribution of gene expression profiling scores of all study patients assigned to the event group and the control group

Fig. 4

The area under the receiving operator characteristics (AUC ROC) for AlloMap score variability to predict future clinical events (death from any cause, re-transplantation or graft failure)